A Trifunctional Linker for Palmitoylation and Peptide and Protein Localization in Biological Membranes

Abstract Attachment of lipophilic groups is an important post‐translational modification of proteins, which involves the coupling of one or more anchors such as fatty acids, isoprenoids, phospholipids, or glycosylphosphatidyl inositols. To study its impact on the membrane partitioning of hydrophobic peptides or proteins, we designed a tyrosine‐based trifunctional linker. The linker allows the facile incorporation of two different functionalities at a cysteine residue in a single step. We determined the effect of the lipid modification on the membrane partitioning of the synthetic α‐helical model peptide WALP with or without here and in all cases below; palmitoyl groups in giant unilamellar vesicles that contain a liquid‐ordered (Lo) and liquid‐disordered (Ld) phase. Introduction of two palmitoyl groups did not alter the localization of the membrane peptides, nor did the membrane thickness or lipid composition. In all cases, the peptide was retained in the Ld phase. These data demonstrate that the Lo domain in model membranes is highly unfavorable for a single membrane‐spanning peptide.


Introduction
The traffic of proteins to the properl ocalization in the cell is necessary for their function. In many instances signals equences determinet he destination of ap rotein, be it the insertion within am embrane or translocation into the lumen of ac ompartmento facell. [1][2][3] Remarkably,c hanges of as ingle amino acid residue can change the localization of lipoproteins from the inner to the outer membrane of Escherichia coli and vice versa. [4] Furthermore, reaching the correct compartment or target membrane is not necessarily enough for proper functioning of the protein. Biological membranes are heterogeneous in structure and localization within as pecific membrane domain has been shown to affect the function of for example, Lymphocyte function-associateda ntigen 1a nd To ll-like receptor 2, which are movedt om ore ordered domains upon bind-ing of as ubstrate. [5][6][7][8][9] Thes ignal outputo fK -Ras changes when al ysine residue is changed into glutamine. [10] This modification alteredt he interaction of K-Ras with anionic lipids and consequently the sorting of those lipids into nanodomains. [10] Protein palmitoylation is ar eversible post-translational modification wherebyo ne or more palmitic acid group(s)a re attachedt oacysteiner esidue (or more seldom, as erine or threonine residue), and this modification has been implicated in the localization of proteins within ag iven membrane. [11][12][13][14] The palmitic acid group changes the hydrophobicity of the complex and may drive its partitioning into as pecific membrane domain. [14] For example, the linker for activation of Tcells (LAT) is enriched in the raft phase of cell-derivedv esicles when palmitoylated. [15] In cells, the doubly palmitoylated H-Ras protein is localized in ad ifferent compartment than the unpalmitoylated K-RAS. [16] In yeast, several amino acid permeases (AAPs) have aC -terminal,a mphipathic a-helix that associates with the inner leaflet of the plasma membrane. In as ubset of these proteins, fore xample the amino acid transporters Gap1 and Tat2, this C-terminal helix is palmitoylated by ap almitoylacyl transferase. [17] The deletion of the amphipathic a-helix does not affect the apparent localizationo fG ap1 and Tat2 but leads to diminished growth on non-fermentable carbon sources. [18] Overall, palmitoylation of membrane proteins is relatively widespreadi nb iology,b ut the functionals ignificance of this modification is in most cases far from clear.
Hydrophobicm ismatch is also known as sorting principle for membrane proteins.W hen the hydrophobic part of am embrane protein or peptide and the lipid membrane have different thickness, the lipids surrounding the protein are distorted which comes with an energetic penalty. [19,20] Proteins preferentially reside in lipid domains with matching thickness, which Attachmento fl ipophilic groups is an important post-translational modificationo fp roteins, which involves the coupling of one or more anchors such as fattya cids, isoprenoids, phospholipids, or glycosylphosphatidyli nositols. To study its impact on the membranep artitioning of hydrophobic peptides or proteins, we designed at yrosine-based trifunctional linker.T he linker allows the facile incorporation of two different functionalities at ac ysteine residue in as ingle step. We determinedt he effect of the lipid modification on the membrane partitioning of the synthetic a-helicalm odel peptide WALP with or without here and in all cases below; palmitoyl groups in giant unilamellar vesicles that contain al iquid-ordered (L o )a nd liquid-disordered (L d )p hase.I ntroduction of two palmitoyl groups did not alter the localization of the membrane peptides, nor did the membrane thickness or lipid composition. In all cases, the peptide was retained in the L d phase.T hese data demonstrate that the L o domain in model membranes is highly unfavorable for a single membrane-spanning peptide.
To study the effect of protein palmitoylation and hydrophobic matching, we designed at rifunctional linker to couple ap eptide or protein to both af luorophore and (a) palmitoyl chain(s).T rifunctional linkersa nd scaffolds are ubiquitousi n chemistry and serve ag reat varietyo fp urposes.E ven the simplest linkersb earing three identicalr eactive groups can be used, taking advantage of stochastic coupling or clever (substoichiometric)i ntroduction of the test molecules. [35][36][37] Trifunctional linkers bearing three different reactive groups are more challenging to synthesize. [38][39][40][41][42] Avery elegant option is am olecule based on at ri-orthogonal "click" scaffold, combining inverse electron demand Diels-Alder (iEDDA) between ac yclooctyne and at etrazine moiety with ac opper-catalyzed alkyneazide click (CuAAC) reactiona sw ell as maleimide coupling. [43] The combination of two CuAAC click reactions with an aldehyde and/ora ctivatede ster coupling is an alternative option. [44] Herein we report the synthesis of at yrosine-based trifunctional linker in which the palmitoyl chains, fluorophore and peptidea re conjugated onto one scaffold via amide, CuAAC and maleimide coupling, respectively.A sm embrane model system to test the partitioning of hydrophobic peptides we used phase-separating giant unilamellar vesicles (GUVs). The chosen lipid compositions separate the GUV membrane into liquid-ordered (L o )a nd liquid-disordered (L d )p hases. [45,46] In addition, we studied the effect of palmitoylationa nd hydrophobic mismatch in this system.W ef ind that the lipid modification of the hydrophobic peptided oes not affect its membrane partitioning and that the L o phase is disfavoredf or all molecules and lipid compositions tested.

Results
Synthesis of tyrosine-based trifunctional linker and characterization of palmitoylated and fluorophore-coupled peptides Amino acids are ag ood starting point for the development of at rifunctional linker,a sm ost naturallyo ccurring amino acids already contain three functional groups. For the trifunctional linker designed here, am aleimide was introducedo nO-propargyltyrosine, followed by an amide coupling with phospholipid 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine (DPPE) to mimic two palmitoyl moieties. The propargyl allows CuAAC with af luorescent dye (Sulfo-Cy3 azide), whilet he maleimide undergoes aM ichael reaction with an introduced cysteine in the WALP23 peptide (sequence:G CGWW(LA) 8 LWWA ; Scheme 1).
We developed as ynthetic route starting from Boc-tyrosine methyl ester (Scheme 2). First, ap ropargyl moietyw as installed on the phenolic-OH after which 1 was obtainedi nh igh yields (Supporting Information). Boc-deprotection followed by amide coupling with 6-maleimidohexanoic acid, using DIC as ac oupling reagent, resulted in amide 3 in moderate yields. Using the Nicolaou ester hydrolysis reaction, [47] employing trimethyl tin hydroxide, the free acid (4)w as obtained in high yields. In the next step, DPPE was introduced on the carboxylic acid in a two-step procedure. First, an activated ester of 4 was generated in situ, followed by rapid removal of the insoluble urea side products.T he activatede ster was added to the solution containing DPPE,r esulting in the phospholipid-modified scaffold 5 in good yield.

Partitioning of the tyrosine-based trifunctional linkers and WALP in phase-separating GUVs
We determined the localization and partitioning of the WALP peptidew ith and without palmitoyl moiety in phase-separating GUVs composed of DPPC, DOPC,a nd cholesterol in 4:3:3 molar ratio ( Figure 1A). This lipid compositionr esults in membraneswith distinct phases,with most of the DPPC and choles-Scheme3.Sequentialconjugation of Sulfo-Cy3 and WALP on scaffolds 4 and 5. terol in the so-called L o phase and most of the DOPC in the L d phase. [48] The constructs WALP-palmitoyl-dye (WPD;c onstruct 9), WALP-dye (WD;c onstruct 8), and palmitoyl-dye (PD;c onstruct 7)w ere used. Each construct was imaged with Atto655-DOPE as marker of the L d phase. The Atto 655 dye displays minimal interaction with the membrane. [49] The Pearson correlation coefficient for the constructs (7, 8 or 9)r elative to the L d marker was calculated to signify the preference of the modified peptide in either of the phases. AP earson's correlationc oefficient of 1i ndicates that the two signals increase and decrease identically,whereas avalue of 0indicates arandom relation between the signals. An egative value would indicate anticorrelation of two signals. We also determinedt he ratio of the molecules for the two phases (L o /L d ), using the meanf luorescence in each phase as am easure of the partitionc oefficient of the constructs.T he palmitoyll ipid (PD) has as light preference for L d (L o /L d ratio of 0.71 AE 0.21), whereas the constructs with WALP (WPDa nd WD)p artition almoste xclusively in L d ( Figure 1B). To eliminate the possibility that the localizationo f WPD and WD in L d is biased by aggregation of the peptide,experiments at 100-fold lower concentration of the functionalized peptides were performed. The resultsw ere the same, and clearly,t he palmitoyl moiety is not sufficient to transfer the WALP peptidef rom the L d to the L o phase ( Figure S1 in the Supporting Information).
Lipid acyl chain saturation, not acyl chain length, determines localization of WALP Next, we investigated the effect of membrane thickness on the partitioning of the constructs.T ot his end, we varied the length of the acyl chains of the lipids, yet maintaining distinct L d and L o phases.W ep repared GUVs composed of cholesterol plus DPPC (16:0)/DOPC (18:1), and GUVs made from cholesterol plus PC with different acyl chain composition: 16:0/16:1, 18:0/18:1, and 18:0/16:1. For these mixtures we assume that the majority of DPPC (16:0) and DSPC (18:0) is in L o and the majority of DOPC and 16:1 PC in L d . [50] Additionally,m ixtures with 14:0, 14:1, and 22:0 PC lipids were tested but conditions resultingi np hase-separation weren ot found. All the constructs co-localize with the L d marker used, which in all cases indicates ap reference for the unsaturated lipid, independent of its acyl chain length (Figure 2).
The localizationo famolecule in the L d or L o phase of a membrane affects its lateral diffusion. Because the mobility of molecules in the L d phase is much faster than in the L o phase , [45] we determined the lateral diffusion coefficient of Atto655-DOPE in GUVs prepared from different lipid compositions. The diffusion coefficients (D) were:1 .53 AE 0.41 (n = 12), 3.11 AE 0.55 (n = 8), 1.10 AE 0.256 (n = 10), and 1.04 AE 0.154 (n = 5) mm 2 s À1 (AE SD) for the 16:0/18:1, 16:0/16:1, 18:0/18:1, and 18:0/16:1 mixtures, respectively.T hese values are consistent with lipid diffusion in the L d phase and at least an order of magnitude faster than what is expected for L o . [45,51] Thus, we concludet hat in all thesel ipid mixtures the Atto655-DOPE localizes in the L d phase of GUVs.
The length of WALP does not affect the membrane localization Next, we increased the length of the WALP peptide to better fit the hydrophobic thickness of the L o phase, which is 0.7 to 1nml arger than in the L d phase. [52][53][54] WALP27 is four amino acids longer than the WALP-23, which resultsi na ni ncreasei n the length of the hydrophobic part of 0.6 nm. [55] We also determined the partitioning of WALP27i nt he presence of up to 5mol %o fG M1, which induces tighter lipid packing and is thought to be am ajor component of rafts in mammalian cells. [56] GM1 is relatively abundant in plasma membranes of the central nervouss ystem of mammals. [57] WALP27 labelled with AlexaFluor 488 (Supporting Information) co-localized with the L d marker DiD in the phase-separating GUVs, irrespective of the presence of GM1 (Figure 3).

Discussion
We report the synthesis and use of at rifunctionall inker to study the localization of membrane proteins and peptides in phase-separating giant unilamellar vesicles (GUVs). The linker allows great flexibility in connecting at riage of molecules such as fluorophores,p eptides (or proteins) and other functionalities, such as lipid moieties. It is possible to selectively modify peptides (or proteins)w ith two different functional groups in a single step. Ours ynthetic approach of couplingt wo palmitoyl groups of DPPE to am embrane peptidev ia the trifunctional linker differs from biological systems where the lipid moieties are attached to cysteineonthe peptide.
WALP and derivative peptides are commonly used as a-helical models of membrane proteins and their interaction with lipid membranes has been studied extensively. [58][59][60] The membrane localization and structure of WALP have been studied in silico and in vitro, using phase-separating GUVs. With both approaches, WALP localizesi nt he L d phaseo ft he membranes. [26,28,61,62] Consistent with these findings, WALP is found in the detergent-soluble fraction of phase-separating large unilamellar vesicles, [63] which is analogoust ot he L d phase of GUVs observedb yo ptical microscopy. [64][65][66][67] We show that the partitioning of the WALP with trifunctional linker is identical to that of genuine membrane peptides. Additionally,W ALP partitions in the L d phase even when two L o -favoringp almitoyl groups are added via the trifunctional linker.T he construct without WALP (PD) is distributed almost equallyb etween L o andL d do-mains.T his result was expected as even GM1, one of the defining componentsofrafts, [5] needs to be complexed with cholera toxin to stain the L o phase specifically. [68] Addingt wo palmitoyl tails to the WALP peptidel owers the energy barrierf or entry into L o phase, but not enough for WALP to localize into the more ordered parts of GUVs with our tertiary lipid compositions. Hydrophobic mismatch created by changing the lipid composition of the membrane or the hydrophobic length of the WALP peptide does not alter the localization or partitioning significantly,i rrespective of the presence of palmitoyl groups.
Palmitoylation hasb een shown to affect the membrane localization of LATa nd hemagglutinin. [69] LATi sl ocalized in the L o phaseofgiant plasma membrane vesicles,a nd the partitioning in L o is diminished after depalmitoylation. [15] However,L AT does not localize in the L o of GUVs composed of DSPC (18:0; 33.3 mol %), DOPC (18:1;3 1.7 mol %), DOPG (18:1; 1.6 mol %) plus cholesterol (33.3 mol %). In addition, palmitoylation of LAT does not affect the partitioning of the molecule in these vesicles, [70] which is consistentw ith our findings on thel ocalization of WALP with or withoutpalmitoyl groups.
What could be the reason for the apparent inconsistency in the membrane domain partitioning of LATa nd other membrane proteins or peptides?T he liquid-ordered and liquid-disordered domains of membranes are qualitative descriptions of lipid ordering,a nd, depending on the actual lipid composition, ad omain can be more, or less, ordered or disordered. We have attempted toa ddress this point by varying the lipid composition of the membrane, whilec onserving microscopically observable L o and L d phases.W eh ave lowered the energy barrier for enteringt he L o domain by creating hydrophobic mismatch, but we alwaysf ind WALP associated with L d .W ec annot rule out that the transfer from L d to L o is more favorable in giant plasma membrane-derived vesicles, consistingo fh undreds of different lipid components, [69] that phase separate with as maller differencei nl ipid order betweenL o and L d . [46] We also note that the angle at which WALP crosses the membrane does not vary much with the membrane thickness. [27,71] In fact, in thin membranes with very high hydrophobic mismatch the peptides are no longeri ncorporatedi nt he membrane rather than highly tilted. [71] Finally,i nv ivo, factorss uch as accessible surface area can also affect the partitioning of membrane peptides and proteins; [72] we have not investigatedt his aspect. GM1, ag anglioside, is often used as raft marker [73][74][75] and associated with the L o phase. [76][77][78] GM1 has been shown to interact with WALP and LAT, therebyf avoring the partitioning of the peptides in the L o phase, at least in coarse grainedmolecular dynamics simulations. [61] All-atoms imulations contradict these findings and show depletion of GM1 near WALP. [79] We note that the latter experiments were carried out in uniform phospholipid bilayersr ather than phase-separating membranes. GM1 could form small nanodomains inside the L o phase, [77,80,81] thereby preventing interaction with WALP.C lusters of GM1 have been found in cell membranes, [80] but also in supported bilayers. [76,77,81] In any case, directinteraction of GM1 and WALP in phase-separating membranes seems unlikely,a t least in our experimental system,b ecause they are spaciously separated with GM1 in the L o phase and WALP in the L d phase. [62,63]

Conclusion
We presentt he design and synthesis of an ew tyrosine-based trifunctional linker,w hich enables conjugation of precious protein with two functional molecules in one step. To show the potentialo fo ur modular platform we study localization and partitioning of membrane-embedded peptides. Contrary to our initial hypothesis, we find in GUVs, prepared from av ariety of lipid mixtures, that ad ouble palmitoyl moiety is not sufficient to change the partitioning of as ingle-membrane spanner like WALP.
For immobilization of GUVs, we used glutaraldehyde from Sigma-Aldrich, product number 340855. APTES ((3-Aminopropyl) triethoxysilane) was obtained from Sigma-Aldrich, product number 440140. WALP (GCGWW(LA) 8 LWWA)w as purchased from Bachem. Sulfo-Cyanine3 azide (Sulfo-Cy3) was purchased from Lumiprobe. All other chemicals were reagent grade and obtained from various commercial sources. High precision coverslips (type #1.5H) were obtained from Ibidi GmbH (product no. 10812). Reactions were monitored by TLC Silica 60 (Merck Millipore), examined under UV (365 nm and 254 nm), and stained by KMnO 4 ,n inhydrin, vanillin or H 2 SO 4 in MeOH (1 %). Flash chromatography was performed on Silica gel 60 (0.040-0.063 mm) from Merck Millipore. 1 HNMR spectra were recorded at 300 or 400 MHz and 13 CNMR spectra were recorded at 75 MHz. The chemical shifts are reported in ppm relative to the residual solvent peak (CDCl 3 at d H = 7.26 ppm, d C = 77.16 ppm). Yields of the dye constructs were based on UV absorption at 548 nm and the molar absorptivity coefficient of Sulfo-Cy3 azide (e = 162 000 m À1 cm À1 at 548 nm), not compensating for the presence of lipid or linker (assumed to not absorb in that region). HPLC was performed on aS himadzu HPLC system equipped with LC-20AD solvent chromatographs, aD GU-20A3 degasser unit, a SPD-M20A PDA detector,aS IL-20A autosampler,aCTO-20A column oven, aC BM-20A system controller and aF RC-10A fraction collector.L C-MS analysis was performed on aW aters Acquity UPLC with TQD mass detector (ESI). High-resolution mass spectra (ESI) were recorded on an Orbitrap XL (Thermo Fisher Scientific).
Immobilization of GUVs: GUVs were immobilized on slides modified with APTES-glutaraldehyde as described previously. [51] In short, coverslips were cleaned in KOH and plasma cleaned. They were then modified in 2% APTES solution for 10 seconds and stored in vacuum until the day of the experiment but always within 48 h. Prior to the experiment, slides were incubated with 5% glutaraldehyde for 30 min, after which the glutaraldehyde was washed away with ddH 2 O. Subsequently,t he GUVs, diluted 10 times in 100 mm NaCl, were placed on the slide.
Imaging: GUVs were imaged on the Zeiss LSM 710 confocal microscope with a4 0x C-Apochromat Corr M27 with NA 1.2 water im-mersion objective. ATTO 655 DOPE and DiD were excited with a 633 nm HeNe laser;t he Sulfo-Cy3 coupled to the WALP peptide was excited with a5 43 nm HeNe laser.Z -stack images of immobilized GUVs were taken. The images of both channels were taken separately to avoid cross talk.
Image analysis: The GUVs were automatically identified by detecting circles in the image. The detection of circles was done using a circle Hough transform [82] for ar ange of radii, resulting in as tack of images;o ne image for each radius. We then created am aximum intensity projection where each pixel contains the maximum value over all images in the stack at the particular pixel location. On this projection we detect peaks by repeatedly finding the brightest pixel, which gives us the centers of the detected GUVs. The radius of each GUV corresponds to the radius on which the maximum pixel value was found when we created the maximum intensity projection ( Figure S2).
After detection of the circles on the image, the non-phase-separating GUVs were filtered out. We calculated the Pearson's correlation coefficient for the fluorescent profile of the L d marker relative to the same profile that was smoothened. Smoothening was done with am oving average filter.Ah igh correlation indicates two separate phases, while low correlation is consistent with as ingle-phase vesicle. We used the correlation of 0.9 as ac ut off for phase-separating vesicles. After the automatic detection all selections were manually inspected and false positives were removed. We used the smoothened profile of the L d marker to classify each intensity value as being either in the L d or L o phase based. Next, the correlation between the L d marker and target construct, and the ratio of the mean intensity of the fluorescence in the L o and L d phases, were calculated on the non-smoothened profiles (an overview of the data analysis is presented in Figure S3).
FRAP experiments: FRAP measurements were performed by imaging as mall area of the membrane of the GUVs to achieve an acquisition time below 40 ms. As pot with ad iameter of 1 mmw as bleached at high laser intensity,a fter which the attenuated laser was used to record images every 40 ms for 6s;t he pre-bleaching fluorescence was obtained from five images prior to the bleach. The halftime of recovery and lateral diffusion coefficients were calculated as described previously, [83] which is based on work of Axelrod and colleagues. [84]